Approaches to HTML table generation

This ticket is to gather requirements and solicit ideas on drawing HTML tables from Vega, possibly with a new html-table oriented renderer, possibly as a Vega plugin.

Goal

Allow users to generate HTML tables from the data after being processed by the Vega transformations, including interactivity.

Approaches

One possible approach is to have a new cell mark. Having this mark automatically creates an HTML table. The table could be positioned either underneath the <canvas> or <svg> element, or on top of it (TBD - multiple layers?). Multiple cell marks could target either one or multiple output tables. Using groups it may be possible to draw tables inside other tables.

Sizing

Table should occupy the entire space of the container (autosize=fit), or be the same size as the graph (autosize=none and pad).

Parameters

• name – ID of the mark
• table – ID of the table - multiple marks could have the same table ID
• column and row (instead of x,y) - integer cell position within a table. The table will dynamically grow whenever the column or row values exceed currently allocated space.
• rowspan and colspan – same as HTML table’s params.
• From the text mark: align, baseline, dir, ellipsis, font, fontSize, fontWeight, fontStyle, limit, text
• border – TBD

Open questions

• Table-level parameters - should there be a top-level and group-level tables section, similar to data? Table nesting based on groups?
• Multiple tables vs a single table - should there always be a single table (similar to how width/height always refer to the main drawing area), or should it be possible to draw multiple ones, and allow table positioning?
• specifying headers/footers - could be done by a separate cell mark targeting the same table and using bold font

[farhanchaudhry]

This would be a great feature. waiting for it.

[iliatimofeev]

@domoritz do you mean something like this?

So vega-embed will have not only canvas but tables section?